Differential DNA Methylation Regions in Cytokine and ... - PLOS

Differential DNA Methylation Regions in Cytokine and Transcription Factor Genomic Loci Associate with Childhood Physical Aggression Nadine Provençal1,2,3, Matthew J. Suderman1,3,6, Doretta Caramaschi2, Dongsha Wang1,2, Michael Hallett6, Frank Vitaro2,7, Richard E. Tremblay2,4,5*., Moshe Szyf1,3*. 1 Department of Pharmacology and Therapeutics, McGill University, Montreal, Quebec, Canada, 2 Research Unit on Children’s Psycho-Social Maladjustment and SteJustine Hospital Research Center, University of Montreal, Montreal, Quebec, Canada, 3 Sackler Program for Epigenetics and Psychobiology, McGill University, Montreal, Quebec, Canada, 4 Department of Psychology and Pediatrics, University of Montreal, Montreal, Quebec, Canada, 5 School of Public Health and Population Science, University College Dublin, Dublin, Ireland, 6 McGill Centre for Bioinformatics, McGill University, Montreal, Quebec, Canada, 7 School of Psycho-Education, University of Montreal, Montreal, Quebec, Canada

Abstract Background: Animal and human studies suggest that inflammation is associated with behavioral disorders including aggression. We have recently shown that physical aggression of boys during childhood is strongly associated with reduced plasma levels of cytokines IL-1a, IL-4, IL-6, IL-8 and IL-10, later in early adulthood. This study tests the hypothesis that there is an association between differential DNA methylation regions in cytokine genes in T cells and monocytes DNA in adult subjects and a trajectory of physical aggression from childhood to adolescence. Methodology/Principal Findings: We compared the methylation profiles of the entire genomic loci encompassing the IL1a, IL-6, IL-4, IL-10 and IL-8 and three of their regulatory transcription factors (TF) NFkB1, NFAT5 and STAT6 genes in adult males on a chronic physical aggression trajectory (CPA) and males with the same background who followed a normal physical aggression trajectory (control group) from childhood to adolescence. We used the method of methylated DNA immunoprecipitation with comprehensive cytokine gene loci and TF loci microarray hybridization, statistical analysis and false discovery rate correction. We found differentially methylated regions to associate with CPA in both the cytokine loci as well as in their transcription factors loci analyzed. Some of these differentially methylated regions were located in known regulatory regions whereas others, to our knowledge, were previously unknown as regulatory areas. However, using the ENCODE database, we were able to identify key regulatory elements in many of these regions that indicate that they might be involved in the regulation of cytokine expression. Conclusions: We provide here the first evidence for an association between differential DNA methylation in cytokines and their regulators in T cells and monocytes and male physical aggression. Citation: Provençal N, Suderman MJ, Caramaschi D, Wang D, Hallett M, et al. (2013) Differential DNA Methylation Regions in Cytokine and Transcription Factor Genomic Loci Associate with Childhood Physical Aggression. PLoS ONE 8(8): e71691. doi:10.1371/journal.pone.0071691 Editor: James D. Clelland, The Nathan Kline Institute, United States of America Received January 16, 2013; Accepted July 2, 2013; Published August 19, 2013 Copyright: ß 2013 Provençal et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: This work was supported by a fellowship from the Genes, Environment and Health Training Program from CIHR to NP, grants from the Canadian Institutes of Health Research, the Social Sciences Humanities Research Council of Canada, the Quebec Health Research Fund (FRSQ) and the Quebec Social Science and Culture Fund (FQRSC) to RET, grants from the Canadian Institutes of Health Research, the Sackler Program in Psychobiology and Epigenetics at McGill University and from the Canadian Institute for Advanced Research to MS. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: The authors have declared that no competing interests exist. * E-mail: [email protected] (RET); [email protected] (MS) . These authors contributed equally to this work.

[5–7]. These children tend to be impulsive, hyperactive, oppositional and rejected by their peers, they also tend to fail in school and have serious social adjustment problems during adulthood [8– 12]. There is good evidence that the parents of children on a high trajectory of physical aggression exhibit similar behavioral problems creating early childhood family environments which do not support learning to regulate physically aggressive reactions [5,7,13–16]. A growing body of research suggests that inflammatory cytokines might have systemic effects in addition to their role in the immune response. Recent studies have shown that changes in

Introduction Physical violence is an important health problem, especially among males [1]. The development of physical aggression has been studied with large population based longitudinal studies from birth to adulthood. Results show that children start using physical aggressions by the end of the first year after birth, increase their frequency from 2 to 4 years of age [2–5], and then reduce the frequency from school entry to adulthood [6]. However, a minority of children (4–7%), mainly males, maintain a high frequency of physical aggression from childhood to adolescence

PLOS ONE | www.plosone.org

1

August 2013 | Volume 8 | Issue 8 | e71691

DNA Methylation in Cytokine Genes and Aggression

cytokine expression levels are associated with various behavioral disorders such as anxiety, depression, suicide, childhood mood disorder and post-traumatic stress disorder (PTSD) [17–28]. In normal men, assessments of hostility, physical aggression, and verbal aggression were positively associated with lipopolysaccharide stimulated TNF-a expression in blood monocytes [29]. Moderate to severe maltreatment during childhood was also observed to be positively correlated with overall change in stressinduced IL-6 concentrations [30]. Other studies examined the association between cytokines and aggression in animals. Gene knockout depletion of IL-6 (2/2) in mice resulted in increased aggression compared to control mice and over-expression of IL-6 in the brain of normal mice increases affiliative behavior [31]. We have recently shown that consistent with these data in mice, physical aggression of boys during childhood is associated with reduced plasma levels of cytokines later in early adulthood [32]. Compared to the control group, men on a chronic physical aggression trajectory from childhood to adolescence had consistently lower plasma levels of five cytokines: pro-inflammatory interleukins IL-1a and IL-6, anti-inflammatory interleukin IL-4 and IL-10, and chemokine IL-8. However, the mechanisms that differentially regulate cytokine expression in white blood cells in chronically aggressive humans are unknown. DNA methylation is involved in programming cell type specific gene expression during development [33]. Consistent with this developmental role of DNA methylation, it is involved in naive CD4+ T cells differentiation into Th1 and Th2 cells [34,35]. The Th2 cytokine locus (IL-4-IL-13-Rad50-IL-5 locus) expressed in Th2 and the IFNc locus expressed in Th1, undergo chromatin remodeling and DNA demethylation during differentiation ([35] and [36] for review). DNA methylation regulates cytokine gene expression (IL-1a [37], IL-6 [38], IL-8 [39], IL-10 [40] and IL-4 [36]) as well as the expression of the transcription factors (TF) that regulate cytokine expression (NFAT5 [41], STAT6 [42] and STAT1 [43]). TFs are also involved in epigenetic reprogramming of their cytokine targets. For example, STATs are required for maintenance of histone acetylation states in the IL-4 and IFNc locus [44] and establishment of histone acetylation and DNA demethylation in IL-4 locus requires the presence of STAT6 [44,45]. Evidence is emerging that in addition to its role in regulating gene expression during differentiation, the DNA methylation pattern is responsive to external environmental exposures including the social environment [46] in animals [47–54] and in humans [55–59]. Importantly, DNA methylation alterations associated with social exposures are not restricted to the brain and are detected in whole blood cell (WBC) DNA [27,55,57,58,60–67]. It is therefore plausible that alteration in DNA methylation of cytokine regulatory regions and the transcription factors that regulate them occur in response to social signals and play a role in human behavior, including aggression. Previous studies from our group and others that have examined associations of genome wide DNA methylation profiles with behavioral exposures have pointed to immune pathways, particularly cytokines. We have shown that maternal deprivation in rhesus macaques is associated with changes in DNA methylation in promoters regulating genes in the immune response pathways [68]. Specifically, transcription factors, cytokines and their receptors, such as NFkB2, IL-16 and IFNcR2 gene promoters, were identified to be differentially methylated in T cells. In humans, we have described association between early life socioeconomic position (SEP) and adult WBC DNA methylation signatures in chemokines (CCL3, CXCL9, CCR1, CCR10, CCR3, CCR6), interleukins (IL-2, IL-4, IL-8, IL1RN), cytokine PLOS ONE | www.plosone.org

receptors (IL10RB, IL6RB) and transcription factors regulating cytokine gene expression (NFATC1, NF-IL3, IkBb) [60]. DNA methylation alterations in white blood cell DNA that were associated with PTSD, included genes involved in inflammation, such as CLEC9A, ANXA2, TLR8 [27]. Lastly, in subjects with a lifetime history of depression, IL-6 and CRP serum levels were found to be elevated, where IL-6 methylation in WBC DNA showed an inverse correlation with circulating IL-6 and CRP [69]. We have previously reported lower plasma cytokines levels in aggressive subjects [32]. We therefore compared the methylation profiles of IL-1a, IL-6, IL-4, IL-10 and IL-8 gene loci and three of their transcription factors NFkB1, NFAT5 and STAT6 gene loci in adult males on a chronic physical aggression trajectory (CPA) between 6 and 15 years of age and males with the same background who followed a normal physical aggression trajectory (control group). Although the promoter area and the immediate 59 regulatory regions play an important role in regulation of gene expression, it is also clear that methylation-dependent regulatory sequences could be found in other parts of the body of the gene or its flanking sequences. Most genome-wide DNA methylation screens focus on promoter regions for reasons of feasibility. By focusing on a short-list of selected genes we were able to examine the entire genomic loci of these genes using a combination of methylated DNA immunoprecipitation (MeDIP) and comprehensive high-density oligonucleotide arrays. Our study provides evidence for association between male physical aggression and differentially methylated regions in cytokines and their transcription factor genomic loci.

Results Associations of Differential DNA Methylation with Physical Aggression in Cytokine Genes We generated and analyzed DNA methylation profiles covering the genomic loci of five cytokines (IL-1a, IL-6, IL-8, IL-4 and IL10) and three transcription factors (NFkB, NFAT5 and STAT6) known to regulate these cytokines. Profiles were generated from T cells and monocytes of adult males with normal (control) and chronic physical aggressive (CPA) trajectories from childhood to late adolescence. The five cytokines were selected for profiling because they have been shown to have lower plasma levels in CPA adult males [32]. As describe above, previous studies have shown that WBCs DNA methylation is associated with environmental exposure and correlate with circulating cytokine levels [61,63,70]. Since WBCs comprise heterogeneous cell populations, which could mask cell-type specific methylation differences, we isolated and examined CD3+ T cells and monocytes since these cell types express the cytokines and transcription factors that were previously shown to be different in the CPA group. The profiles were generated using MeDIP followed by hybridization to microarrays with probes placed at 100 bp-spacing starting from about 240 Kb before the transcription start site of each gene of interest to about 40 Kb after its transcription end site. Thus, the profiles included not only the promoters and bodies of these genes but also more distant sites in order to identify any genomic site with methylation levels altered in CPA individuals that might explain the previously observed lower cytokine levels. Differences between the CPA and control groups were found throughout each cytokine loci in both cell types analyzed (Table 1). In total, 48 differentially methylated regions associated with CPA (,1000 bp, log2-fold .0.5, FDR ,0.2) were identified in the cytokine loci, 20 in T cells and 28 in monocytes. Surprisingly, none of these were located in the promoters of the candidate cytokines, defined as the region from 22000 bp to +1000 bp of the transcription start site, although one 2



differentially methylated region was found just outside the promoter (22042 bp) of IL-6 (Figure 1). Two differentially methylated regions were found in the enhancer/insulator locus control region (LCR) of the IL-4 and IL-10 loci (Figure 2). Each cytokine gene loci included several other genes, and some of these non-candidates were in fact closer to several differentially methylated regions than the originally short-listed genes. Tables listing all differentially methylated regions associated with aggression for each cell type along with the gene nearest to the region can be found in Spreadsheet S1. To better understand the regulatory factors potentially affected by the differentially methylated regions associated with aggression, we compared each differentially methylated region against transcription factor binding sites, DNase clusters and histone mark enrichments recorded in the Encyclopedia of DNA Elements (ENCODE) database (http://genome.ucsc.edu/ENCODE/), de-

rived previously from 24 different cell lines (histone mark profiles were available for only 8 of the cell lines). In total, 25 of the 48 differentially methylated regions associated with CPA coincided with transcription factor binding sites (Table 1), and many of the bound transcription factors are known regulators of cytokines such as NFkB, GR, PU.1, STAT and BAFF family members. A high fraction of the differentially methylated regions associated with aggression (35 of 48) coincided with DNase hypersensitive clusters indicating that these regions have an open chromatin state in at least one cell line and therefore likely to play regulatory roles (Table 1). The number of cell lines in which the DNase clusters were observed is shown in Figures 1, 2, 3 as well as in the Figures S1, S2, S3. Histone marks are included in Figures 1, 2, 3 and Figures S1, S2, S3 but only those delineated in cell lines related to T cell and monocytes: a lymphocyte cell line (GM12878), a lymphoblastoid cell line, and K562, an erythroleukemia cell line.

Figure 1. DNA methylation differences between CPA (n = 8) and control (n = 12) groups in pro-inflammatory cytokines IL-6 and IL-1a loci in T cells. Expanded views from the UCSC genome browser of IL-6 (A) and IL-1a (B) loci located on chromosomes 7 and 2 are depicted (top panel). In both panels, the first track shows the average MeDIP probe log2-fold differences (scale top panel: 20.4 to 0.4, botom panel: 21.0 to 1.0) between chronic physical aggressive (CPA) and control groups for T cells. In black are probes that are more methylated and in gray are those that are less methylated in the CPA group. Highlighted in blue are regions significantly differentially methylated between the groups. The second track shows Pearson’s correlation coefficient values (scale top panel: 20.4 to 0.4, botom panel: 20.7 to 0.7) calculated between the MeDIP microarray probe intensities and the plasma IL-6 (A) and IL-1a (B) levels obtained from the same subjects (n = 20). In red are probes whose methylation levels correlate positively with the cytokine level in plasma and in green are those that correlated negatively. In the top panel, the lowest track shows the average methylation level for all the subjects in T cells estimated from the microarray data. The bottom panel zooms on the closest region upstream of the TSS where the CPA group is found significantly less methylated than the control group in the IL-6 (A) and IL-1a (B) loci. In both regions, an overall positive correlation between methylation and cytokine level in plasma is observed but only reached significance after correcting for multiple testing in IL-1a (B) differentially methylated region. The regulatory elements from ENCODE identified in these regions (see methods) are shown in the additional tracks. First, shown with black lines, is the location of individual CpG sites. Second, is the location of DNase hypersensitive clusters where black indicate strong signal and grey a weaker signal from ChIP-seq data in 24 cell lines. Third is the location of transcription factors (TF) identified from ChIP-seq data in 24 cell lines where black indicate a strong and grey weaker signal occupancy. The letter next to the TF boxes identified the cell line where it was found enriched (see Table S1 for the full legend). The last tracks, identified the level of enrichment of three histone marks determined from ChIP-seq assay, histone 3 lysine 4 tri- and mono-methylation as well as histone 3 lysine 27 acetylation in two cell lines, GM12878 (pink) and K562 [106]. doi:10.1371/journal.pone.0071691.g001


3



Figure 2. DNA methylation differences between CPA (n = 8) and control (n = 12) groups in anti-inflammatory cytokines IL-4 and IL10 loci in T cells. Expanded views from the UCSC genome browser of IL-4 (A) and IL-10 (B) loci located on chromosomes 5 and 1 are depicted (top panel). In both panels, the first track shows the average MeDIP probe log2-fold differences (scale top panel: 20.4 to 0.4, botom panel: 21.0 to 1.0) between chronic physical aggressive (CPA) and control groups are shown for T cells. In black are probes that are more methylated and in gray are those that are less methylated in the CPA group. Highlighted in blue are regions significantly differentially methylated between the groups in T cells for IL-4 (A) and IL-10 (B). The second track shows Pearson’s correlation coefficient (scale top panel: 20.4 to 0.4, botom panel: 20.7 to 0.7) calculated between the MeDIP microarray probe intensities and the plasma IL-4 (A) and IL-10 (B) levels obtained from the same subject (n = 20). In red are probes whose methylation levels correlate positively with the cytokine level in plasma and in green are those that correlated negatively. In the top panel, the last track shows the average methylation level for all the subjects in T cells for IL-4 (A) and IL-10 (B) estimated from the microarray data. The bottom panel shows differentially methylated regions located within known Th2 LCRs where the CPA group is found significantly less methylated than the control group in the IL-4 locus (A) and significantly more methylated in the IL-10 locus (B). For the first differentially methylated region in IL4, an overall positive correlation between methylation and IL-4 level in plasma is observed but did not reach significance after correcting for multiple testing (A, bottom panel). In the region more methylated in the CPA group of the IL-10 locus (B, bottom panel), the overall correlation with IL-10 level in plasma is negative although not significant after corrections for multiple testing. The regulatory elements from ENCODE identified in these regions (see methods) are shown in the additional tracks. First, shown with black lines, is the location of individual CpG sites. Second, is the location of DNase hypersensitive clusters where black indicate strong signal and grey a weaker signal from ChIP-seq data in 24 cell lines. Third is the location of transcription factors (TF) identified from ChIP-seq data in 24 cell lines where black indicate a strong and grey weaker signal occupancy. The letter next to the TF boxes identified the cell line where it was found enriched (see Table S1 for the full legend). The last tracks, identified the level of enrichment of three histone marks determined from ChIP-seq assay, histone 3 lysine 4 tri- and mono-methylation as well as histone 3 lysine 27 acetylation in two cell lines, GM12878 (pink) and K562 [106]. doi:10.1371/journal.pone.0071691.g002

In several examples enrichment of particular histone marks coincides with differential methylation associated with CPA. The H3K4Me1 histone mark is mainly associated with enhancers and with DNA regions downstream of transcription start sites whereas H3K4Me3 is mainly associated with promoters and enhancers that are active or poised to be activated [71]. The H3K27Ac mark is thought to enhance transcription possibly by blocking the spread of the repressive histone mark H3K27Me3 [72].

DNA methylation and cytokine levels. Of the 48 differentially methylated regions associated with aggression, 7 regions were significantly correlated with cytokine levels encoded by the closest genes in blood plasma (FDR ,0.2; Table 1). All but two were positively correlated with the cytokine levels in blood. The differentially methylated region associated with aggression closest to a cytokine transcription start site was located 2042 bp upstream of the pro-inflammatory cytokine IL-6 and was less methylated in the CPA group compared to controls in T cells (Figure 1A and Figure S1). This region contains a DNAse hypersensitive site and a peak of H3K4Me1 in the two lymphocytes cell lines that were studied in the ENCODE project. Interestingly, this region also coincides with the binding sites of

Associations between Differentially Methylated Regions and Cytokine Levels Using previously measured cytokine protein levels in the same set of males [32], we were able to test for associations between PLOS ONE | www.plosone.org

4



5

IL-4

IL-8

IL-1a

IL-6

Gene locus

Control

35

0.52 (0.14)

CPA

+13.5

Control Control

216.5

+3 (Intron 2)

231.5 Th2 LCR

Control

Control

Th2 LCR

Control

235

Control

36

32

CPA

Control

8

17

CPA

216.5

CPA

0.6 (0.06)

c-Jun, BAF155, BAF170, Brg1, Ini1, NFkB, CTCF, GR, TAF-1, JunD

PU.1, Rad21, NFkB, CTCF, EBF, BCL3

NFKB, BCL3, SP1, EBF, IRF4, POU2F2, BCL11A, EBF, p300, BATF

NFkB

c-Jun, GR

PU.1

–

–

–

–

c-Jun, JunD

–

219

–

CPA 20.53 (0.07)

– c-Myc, BAF155, STAT1, JunD

CPA

CPA

227

CPA

229.5

236.5

240

–

– 0.8 (0.1)

CPA

260.5

20.58 (0.02)

–

CPA

249.5 Control

NFkB, EBF, Pbx3, BCL11A, TCF12

Control

246.5

Control

–

Control

244.5

281

–

Control

240

NFkB, Max, C-Jun, FOSL2, EBF, SP1, BATF, IRF4, BCL11A, JunD, Pax5, TAF1, TCF12, Rad21, c-Fos, PU.1

CEBPB

Control

Control

224

226

PU.1 –

Control

Pbx3

–

NRSF

Control 0.45 (0.07)

0.5 (0.08)

–

NFkB, PU.1

7

Control

CPA

Control

Control

Control

Transcription factors (ENCODE)

19

Control

21

+33.5

CPA

T cells

Monocytes

Monocytes

T cells

Pearson’s correlation coefficient R (FDR)

More methylated in

18

24

Distance from TSS (kb)

Table 1. Differentially methylated regions between CPA (n = 8) and control (n = 12) groups and correlation with cytokine levels in plasma.

yes

yes

yes

yes

yes

yes

yes

yes

–

yes

yes

yes

–

yes

yes

–

yes

–

yes

yes

–

yes

yes

–

yes

yes

yes

yes

yes

yes

DNase cluster (ENCODE)




6

CPA

27

Control

7

–

yes

yes

–

–

yes

–

–

yes

DNase cluster (ENCODE)

NfkB, BAF170 NFkB, CTCF, CEBPB, JunD, RAD21, STAT1

CPA

237

yes

yes

yes

BCL3, Pbx3 CPA

Control

216

223

NFkB, EBF, BATF, BCL11A, TCF12, Max, IRF4, BCL3, yes PAX5, SP1

CPA

Pol2, miRNA, NFkB, ZNF263, IRF4, BCL11A, yes PU.1, EBF, BATF, PAX5, SP1, POU2F2, JunD, FOSL2, Rd21

–

GR

PU.1

–

–

SREBP1

–

SREBP1

NFkB, SREBP2


4

CPA

CPA

42

6 Th2 LCR

Control

Control

37

Control

43

CPA

+34.5

+31.5

Control CPA

+23.5

T cells

Monocytes

Monocytes

T cells

Pearson’s correlation coefficient R (FDR)

More methylated in

+8 (Intron 3)


In bold are regions that are illustrated in Figures 1, 2 and 3. doi:10.1371/journal.pone.0071691.t001

IL-8

Gene locus

Table 1. Cont.




Figure 3. DNA methylation differences between CPA (n = 8) and control (n = 12) groups in pro-inflammatory chemokine IL-8 locus in monocytes. Expanded views from the UCSC genome browser of IL-8 (A) locus located on chromosomes 4 is depicted (top panel). In both panels, the first track shows the average MeDIP probe log2-fold differences (scale top panel: 20.4 to 0.4, botom panel: 21.0 to 1.0) between chronic physical aggressive (CPA) and control groups are shown for monocytes. In black are probes that are more methylated and in gray are those that are less methylated in the CPA group. Highlighted in red are regions significantly differentially methylated between the groups. The second track shows Pearson’s correlation coefficient (scale top panel: 20.4 to 0.4, botom panel: 20.5 to 0.5) calculated between MeDIP microarray probe intensities (DNA methylation levels) and the plasma IL-8 levels obtained from the same subject (n = 20). In red are probes whose methylation levels correlate positively with the cytokine levels in plasma and in green are those that correlated negatively. In the top panel, the last track shows the average methylation level for all the subjects in monocytes estimated from the microarray data. The bottom panel shows the closest region upstream of the TSS where the CPA group is found significantly less methylated than the control group in the IL-8 locus. In both regions, an overall negative correlation between methylation and cytokine level in plasma is observed but did not reach significance after correcting for multiple testing. The regulatory elements from ENCODE identified in these regions (see methods) are shown in the additional tracks. First, shown with black lines, is the location of individual CpG sites. Second, is the location of DNase hypersensitive clusters where black indicate strong signal and grey a weaker signal from ChIP-seq data in 24 cell lines. Third is the location of transcription factors (TF) identified from ChIP-seq data in 24 cell lines where black indicate a strong and grey weaker signal occupancy. The letter next to the TF boxes identified the cell line where it was found enriched (see Table S1 for the full legend). The last tracks, identified the level of enrichment of three histone marks determined from ChIP-seq assay, histone 3 lysine 4 tri- and mono-methylation as well as histone 3 lysine 27 acetylation in two cell lines, GM12878 (pink) and K562 [106]. doi:10.1371/journal.pone.0071691.g003

scription factor-binding sites, including those of both activators and repressors. Transcription factors BATF, NFkB and PU.1 have the strongest binding signal in this region. In the locus of IL-4, we identified ten regions that were differentially methylated between CPA and controls, three of these regions were found in T cells. Two of these regions that are less methylated in the T cells of CPA individuals, contain several ENCODE regulatory elements (Figure 2A bottom, Figure S2A), including a DNase hypersensitive cluster, enrichment for histone marks H3K4Me3 and H3K27Ac, and several transcription factor binding sites (Figure 2A bottom, highlighted in blue). Theses regions are located in the Th2 cytokine Locus Control Region (LCR) known to regulate the expression of IL-4 locus.

transcription factors PU.1 and NFkB. PU.1 is known to act as both a repressor and activator of several cytokine genes [73] and has been proposed to be involved IL-6 expression [74]. Moreover, this IL-6 upstream region has been previously reported to have basic transcriptional activity in gene reporter luciferase assays [75]. A second region less methylated in the T cells of CPA individuals compared to controls was located in the IL-1a locus (Figure 1B and Figure S1B), located 27 Kb upstream of IL-1a. The region is of interest because its methylation levels are significantly correlated with IL-1a levels (R = 0.52; FDR ,0.14), and it contains many regulatory elements identified by ENCODE (Figure 1B bottom, highlighted in blue). These include a DNase hypersensitive site, H3K4Me1 and H3K27Ac enrichment peaks and several tran-


7



Table 2. Differentially methylated regions between CPA (n = 8) and control (n = 12) groups identified in the loci of three transcription factors known to regulate cytokines.

Gene locus


More methylated in Monocytes

232 p150

DNase clusters (ENCODE)

T cells Control

yes

28 p150

CPA

PU.1

yes

+6.5 p150

CPA

–

–

+10 p150

CPA

+12 p150

CPA

+24 p150

CPA

+26 p150

NFKB1 Two transcripts: p150 and p50


CEBPB

yes

–

–

–

yes

Control

BATF, NFkB, EBF

yes

Control

–

yes

+30.5 p150

CPA

+34 p150, 242 p50

CPA

JunD, IRF4, BCL11A, NFkB, BATF, POU2F2, PAX5, C-Jun

yes

+36 p150, 240 p50

CPA

–

yes

+39 p150, 237 p50

CPA

–

–

+53 p150, 221 p50

CPA

–

yes

+65 p150, 29 p50

CPA

NFkB

yes

Control

–

–

Control

–

yes

+76 p150, TSS p50 +80 p150,

CPA

+4 p50 +8 p50

CPA

IRF4, BAF, EBF, JunD

yes

+11.5 p50

Control

–

yes

+20 p50 +26 p50

Control Control

+31 p50 +35 p50

CPA CPA

+37 p50 +45 p50

CPA CPA

+53 p50

CPA

yes

–

–

–

–

–

yes

–

yes

FOSL2, HNF4A, CTCF, USF-1

yes

GR

yes

–

–

–

–

–

–

+55 p50

CPA

+68 p50

Control

+83 p50

Control

223

Control

C-Jun, PAX3, JunD, NFkB

yes

9

CPA

–

–

CPA

–

–

CPA

–

yes

–

yes

22

Control

+25.5

CPA

43

CPA

52

Control

CEBPB

53

CPA

–

54 55 NFAT5

NFkB

CPA

62 CPA

77

CPA

84

Control

85

CPA

87 91


–

yes

CPA

–

–

Control

67

yes

Control

Control

–

–

–

–

–

–

Ini1

–

Control

–

–

CPA

Control

–

–

CPA

Control

–

–

8



Table 2. Cont.

Gene locus


More methylated in


DNase clusters (ENCODE)

Monocytes

T cells

112

CPA

Control

–

–

128

CPA

–

–

136

CPA

–

–

+53 isoform 2

CPA

NFkB

yes

+25 isoform 2

CPA

HEY1, Brg1, TAF1, C-Myc, NFkB, STAT2, Max, BAF155, yes Ini1, STAT1, Pu.1, BAF170

+20 isoform 2

Control

Ini1

yes

+6 exon 8 to 10 isoform 2

Control

–

yes

+3 exon 1 isoform 2

Control

–

yes

STAT6 Two TSS isoform 2, +18 transcripts: isoform 1 isoform 1 and isoform 2 21.5 isoform 1

CPA

C-Myc, max, NFkB, brg1, BAF170, STAT1, yes BAF155, TAF1, Ini1, P1, POUF2L2, BCL3, PU.1, IRF4, PAX5, HEY1, CEBPB, EBF, TCF12, NRSF

CPA

HNF4, CEBPB, FOSL2, RXRA, p300, BHLHE40, JunD, ERRA, USF-1, NFkB

yes

210 isoform 1

Control

HNF4, CEBPB, FOSL2, BAF155, p300, HEY1, RXRA, JunD

yes

214 isoform 1

Control

–

yes

223 isoform 1

Control

–

yes

In bold are regions that are illustrated in Figures 4. doi:10.1371/journal.pone.0071691.t002

Three differentially methylated regions associated with aggression lie within 10 Kb downstream of the IL-10 transcription start site. Two regions are differentially methylated in T cells the first (+6 Kb) is more methylated in CPA individuals whereas the second (+7 Kb) is less methylated. The third difference lies in the 39UTR of IL-10 and is more methylated in monocytes of CPA individuals (Figure 2B and Figure S2B). Only the first region (+6 Kb) contains several of the ENCODE regulatory elements (Figure 2B bottom, highlighted in blue), including a strong DNase hypersensitive cluster, enrichment for histone marks and several transcription factor binding sites. Interestingly, this region was previously found to have promoter/enhancer activity in Th2 cells and was proposed to function as a LCR for the IL-10 gene locus [76]. The IL-8 locus contains eight regions differentially methylated between CPA and controls in monocytes and three in T cells (Figure 3A and Figure S3). Since IL-8 is mainly expressed in monocytes, we focused on two regions more methylated in monocytes of CPA individuals that are 16 Kb and 19 Kb upstream of IL-8, respectively (Figure 3A bottom, highlighted in red). Both regions, though located far from the gene promoter, were found to include a strong DNase hypersensitive cluster, and the second region contained a small H3K4Me1 peak in the K652 cell lines. Interestingly, binding sites of transcriptional activators cJun and JunD were identified in the first region. C-Jun and JunD are both part of the AP-1 transcriptional complex, a well-known activator of many genes including cytokines [77].

classical regulators of the cytokines analyzed above. Differentially methylated regions between the CPA and control groups were identified throughout the loci in both cell types analyzed (Table 2) with both increased and decreased methylation levels in CPA individuals compared to controls. Each locus had at least ten differentially methylated regions (,1000 bp, log2-fold .0.5, FDR ,0.2). Similarly to the cytokine loci, the locations of some differentially methylated regions associated with aggression coincided with the location of regulatory regions such as the promoters (as defined earlier 22000 bp to 1000 bp of TSS) of both STAT6 isoforms as well as in the NFkB1 short transcript (p50) that were more methylated in the CPA group (Table 2). We found 37 differentially methylated regions associated with CPA within the NFkB locus, 21 of these regions were located within the gene body (Figure S4). In the NFAT5 locus, 19 differentially methylated regions between CPA and controls were identified and all of them except one were located within the body of the gene (Figure S5). A total of 10 regions were identified in STAT6 locus, two of the differentially methylated regions associated with CPA were found in T cells and all other differentially methylated regions associated with CPA in STAT6 were found in monocytes (Figure S6). As we did for the cytokine genes, we examined whether the differentially methylated regions associated with aggression contained previously published transcription factor (TF) binding sites, DNase hypersensitive clusters and enrichment for histone modifications H3K4Me1, H3K4Me3 and H3K27Ac using the ENCODE database (http://genome.ucsc.edu/ENCODE/). We found that most of the differentially methylated regions associated with CPA identified in NFkB1 and STAT6 loci in both cell types were located within DNase hypersensitive clusters, transcription factors binding sites and H3K4Me1, H3K4Me3 and H3K27Ac peaks, supporting a regulatory role for these regions (Table 2 and Figures S4 and S6). In contrast, only a few differentially

Associations of DNA Methylation with Physical Aggression in the Loci of Transcriptional Regulators of Differentially Expressed Cytokines The microarray used in this study included the genomic loci of three transcription factors, NFkB1, NFAT5 and STAT6 that are PLOS ONE | www.plosone.org

9



Figure 4. DNA methylation differences between CPA (n = 8) and control (n = 12) groups in cytokine’s transcription factor STAT6 in T cells. Expanded views from the UCSC genome browser of STAT6 (A) locus located on chromosomes 12 is depicted (top panel). In both panels, the first track shows the average MeDIP probe log2-fold differences (scale top panel: 20.4 to 0.4, botom panel: 21.0 to 1.0) between chronic physical aggressive (CPA) and control groups for T cells. In black are probes that are more methylated and in gray are those that are less methylated in the CPA group. Highlighted in blue are regions significantly differentially methylated between the groups. The second track shows Pearson’s correlation coefficient (scale top panel: 20.4 to 0.4, botom panel: 20.6 to 0.6) calculated between MeDIP microarray probe intensities and the plasma IL-4 (first) and IL-10 (second) levels obtained from the same subject. These correlations did not reach significance after correcting for multiple testing in T cells. In red are probes whose methylation levels correlate positively with the cytokine level in plasma and in green are those that correlated negatively. In the top panel, the last track shows the average methylation level for all the subjects in T cells estimated from the microarray data (n = 20). The bottom panel shows the two regions close to the TSS of each STAT6 isoform where the CPA group is found significantly more methylated than the control group. The regulatory elements from ENCODE identified in these regions (see methods) are shown in the additional tracks. First, shown with black lines, is the location of individual CpG sites. Second, is the location of DNase hypersensitive clusters where black indicate strong signal and grey a weaker signal from ChIP-seq data in 24 cell lines. Third is the location of transcription factors (TF) identified from ChIP-seq data in 24 cell lines where black indicate a strong presence and grey weaker signal occupancy. The letter next to the TF boxes identified the cell line where it was found enriched (see Table S1 for the full legend). The last tracks, identified the level of enrichment of three histone marks determined from ChIP-seq assay, histone 3 lysine 4 tri- and mono-methylation as well as histone 3 lysine 27 acetylation in two cell lines, GM12878 (pink) and K562 [106]. doi:10.1371/journal.pone.0071691.g004

We have further confirmed the methylation differences between CPA and control groups and correlation with the cytokine levels in plasma in regions within the IL-6, IL-8 and STAT6 loci by pyrosequencing (Figure 5A, 5B and 5C; see Methods). Intriguingly, we note that the pyrosequencing methylation levels in STAT6 gene are significantly associated with IL-4 levels in plasma.

methylated regions between CPA and controls in NFAT5 locus were located within regulatory elements. None of the regions were located within CpG islands. The methylation profile of the STAT6 locus in T cells is illustrated in Figure 4. Two regions in this locus are more methylated in the T cells of CPA individuals compared to controls (Figure 4A and Figure S6, highlighted in blue). The first region is located at the transcription start site of the STAT6 isoform 2 gene (Figure 4 left bottom panel), and the second region is located 1 Kb upstream of STAT6 isoform 1 (Figure 4 right bottom panel). Both regions coincide with DNase hypersensitive clusters and many transcription factor binding sites such as IRF4, HEY1 and NFkB. Enrichment for H3K4Me3 and H3K27Ac marks were found in the first region (TSS isoform 1) in both cell lines whereas only a small peak of H3K4Me1 enrichment was found in the second region in the GM12878 cell line.


Discussion Several lines of evidence suggest that cytokines are associated with animal and human aggression [78–80] and IL-6 was causally linked to aggression in mice by gene knockout evidence [31]. Our analyses of males whose level of aggression were followed for a 22 year period revealed an association between cytokine levels in plasma and chronic physical aggression: young adult men with a history of chronic physical aggression during childhood have lower baseline concentrations of two pro (IL-1a, IL-6 and IL-8) and two 10



Figure 5. Pyrosequencing analyses of differentially methylated sequences in T cells and monocytes between CPA (n = 8) and control (n = 12) groups. DNA methylation differences (%) between the aggressive groups in the IL-6 (A), IL-8 (B) and STAT6 (C) genes as determined by pyrosequencing. CpG sites near the significant probes were analyzed in triplicate. For each gene, the mean methylation per group per CpG with the standard error of the mean (SEM) are shown in the bar graph (top panel). The rightmost bar indicates the mean methylation levels of all CpG sites analyzed in the region. A map of the sites relative to the transcription start site is shown above the bar graph. Each line with a circle represents a CpG site. The location of probes which fold difference is significantly different between the groups is identified by a grey square where dark and light grey indicated more or less methylated in the CPA group respectively. Red arrows delimit the region analyzed by pyrosequencing. Correlations between mean methylation levels obtained by pyrosequencing for each gene and expression levels in plasma obtained by ELISA for each cytokine [32] is shown in the bottom panel. Correlation statistics were performed using Pearson correlation. All error bars represent standard error of the mean (SEM). The symbol ‘#’ denotes a p-value #0.1 and the symbol ‘*’ denotes a p-value #0.05 as determined by a Student t-test. doi:10.1371/journal.pone.0071691.g005

of cytokine expression. Further work will be needed to investigate the exact role of these new candidate regulatory regions in the regulation of transcription of these genes and particularly in the context of problems with regulation of physical aggression during childhood. Interestingly, four of our differentially methylated regions associated with aggression were located within known cytokines regulatory regions. Particularly, three regions lie within Th2 cytokines LCRs, two in the IL-4 loci and one in the IL-10 loci. These LCRs have been shown to have enhancer/insulator activity in Th2 cell differentiation [44,76,81]. Changes in chromatin remodeling and DNA methylation in this region are involved in regulating expression of IL-4 and IL-13 during Th2 differentiation [44,81]. It was previously proposed that DNA demethylation is

anti (IL-4 and IL-10) -inflammatory cytokines than comparable young adult men without a history of chronic physical aggression during childhood [32]. Here we show data that is consistent with the hypothesis that DNA methylation could be part of the mechanisms responsible for differences in cytokine levels in the chronic aggression group. Indeed, we found differentially methylated regions to associate with CPA in both the cytokine loci as well as in their transcription factors loci analyzed. Some of these differentially methylated regions found in IL-6, IL-4, IL-10, STAT6 and NFkB were located in known regulatory regions whereas others, to our knowledge, were previously unknown as regulatory areas. However, using the ENCODE database, we were able to identify key regulatory elements in many of these regions that indicate that they might be involved in the regulation


11



required for the maintenance of IL-4 LCR enhancer activity in Th2 differentiated cells [81]. STAT6 transcription factor was also shown to be required for the remodeling of this LCR during Th2 cell differentiation [81]. We observed in the same region lower methylation in the CPA individuals in T cells and overall lower IL4 plasma levels compared to controls (Figure 2A). Since we did not examine the methylation levels in specific T cells subpopulations, we don’t know what this demethylation in CPA individuals means in term of Th2 LCR activity. The fact that it occurs in this important regulatory region suggests however that the activity of the IL-4 locus might be altered in individuals with history of chronic physical aggression during childhood. On the other end, we found higher methylation in the promoter of STAT6 genes in the CPA individuals (Table 2). STAT6 is a well-known regulator of IL-4 in T cells, specifically in Th2 differentiating cells. As mentioned earlier, it can induce expression of IL-4 by interacting with the Th2 LCR [81] but also by activating the expression of the transcription factor GATA-3 that bind directly to the IL-4 promoter [82]. The methylation differences observed in the promoter of isoform 2 as well as in intron 1 of isoform 1 were validated by pyrosequencing. Intriguingly, we note that the pyrosequencing methylation levels in STAT6 gene are significantly associated with IL-4 levels in plasma, further supporting a regulatory role for these sites in regulation of IL-4 by STAT6. This is consistent with the hypothesis that increased methylation leads to a decrease in STAT6 expression and that could lead to reduced activation of IL-4 locus and lower levels in plasma. DNA methylation analyses were not previously performed for the LCR in IL-10 locus but it stands to reason that as the LCR in IL-4 locus, DNA demethylation is needed for the maintenance of this LCR functionality. The LCR in IL-10 locus was found more methylated in the CPA individuals compared to controls. These results are consistent with the hypothesis that higher methylation in T cells in this region may account, in part, for the decreased in IL-10 expression observed in the plasma of the CPA group. Similarly, IL-8 results are consistent with the hypothesis that higher methylation in CPA individuals reduces IL-8 expression levels by inhibiting binding of transcriptional activators. In, IL-1a, IL-6 and IL-4 LCRs our study delineated regions with lower methylation in T cells in CPA individuals who also show lower levels of these cytokines in plasma (Figure 1A, 1B and 2A bottom, highlighted in blue). These results might be consistent with the hypothesis that lower methylation in these regions permits more frequent binding of a putative repressor leading to the decreased IL-4, IL-6 and IL-1a expression. Indeed, there is evidence that increased DNA methylation of certain regions is associated with increased gene expression. For example, DNA methylation in a repressor-binding site of the multidrug-resistance gene (MDR-1) enhancer was shown to increase its expression [83]. Another example is the H19/IGF2R imprinted control regions (ICR). Methylation of the ICR on the paternal allele prevents H19 activity but allows IGF2 activity whereas hypomethylation on the maternal allele allows H19 activity but prevents IGF2 activity [84]. These hypotheses on the effect of DNA methylation on cytokine genes expression need to be tested in vitro as well as in vivo as the data presented here only reveal associations but do not provide causal relationships between the differentially methylated regions and expression. Our results also point to the value of locus-wide analysis of DNA methylation. Most studies are limited to promoter regions. Had we used this approach here we would have missed most or all of the differentially methylated regions that we had identified in this study. Many of which contain high concentration of regulatory signals. This suggests that many of the published failures in PLOS ONE | www.plosone.org

identification of differentially methylated regulatory regions might be a consequence of limiting the analyses to known promoter regions. Taken together the fact that a high fraction of the differentially methylated regions between the CPA and control groups were localized in either known regulatory regions or regions that were independently shown to contain a high concentration of regulatory chromatin marks, DNase hypersensitivity and transcription factors, supports the hypothesis that they are involved in differential regulation of cytokine expression between CPA and control groups. Indeed, several methylated regions located within the cytokine loci were found to significantly correlate with cytokine levels in plasma. Although our list of differentially methylated regions associated with CPA is comprehensive it is still possible that additional differentially methylated regulatory regions exist in these loci that were not discovered in our analysis. The small sample size could result in loss of power to detect additional significant correlations. In addition, the associations of differentially methylated regions and cytokine expression delineated here are probably an underestimation since we have used plasma cytokine levels rather than transcription rates as a measure of cytokine expression. There are many steps in between initiation of transcription at the promoter and cytokine levels in plasma. Unfortunately, the study design did not allow us to isolate high quality RNA because it was difficult to request that the subjects come to the lab for their blood perfusion, particularly with the chronic aggression group, resulting in a delay between cell harvest and RNA processing. In addition, DNA methylation and expression are fundamentally different measures of cell functional states. DNA methylation levels directly correspond to the number of cells with a methylated/unmethylated cytosine (as a state of methylation of an allele in a particular site is a binary count, in a given cell a site is either methylated or not) whereas protein and mRNA levels are analog measures that average the number of mRNA and protein molecules produced by the cell population. A low level of mRNA or protein measure could indicate either high level of expression in a few cells or low level of expression in many cells. Several methodological considerations arise in our study. First, with blood available only in adulthood, we cannot establish when during development chronic physical aggression became associated with these particular DNA methylation alterations and cytokine level reduction in plasma. Based on the longitudinal data obtained from this cohort, we can expect that DNA methylation changes appear early in life as chronic physical aggression trajectories usually start in infancy and are associated with numerous factors during the prenatal and early postnatal periods [5]. However, it stands to reason that there is a developmental trajectory of these DNA methylation changes and that changes seen in adults will not be identical to the early changes seen in aggressive children. The second issue relates to cause and effect. We have demonstrated here a highly significant association between DNA methylation alterations and CPA. These changes could be either a cause or an effect of CPA life style. The issue of causality is particularly difficult in human studies although a parallel longitudinal analysis of behavioral and DNA methylation changes could have established at least the temporal relationship between differential DNA methylation of cytokines and aggressive behavior. In addition, childhood abuse or neglect is also known to increase the risk of aggression in adolescents and adults [85] and was also found to associate with DNA methylation differences [59]. Therefore, it is possible that child abuse acts as a third factor in explaining the reported association between aggression and DNA methylation.

12


mother and father occupational score, familial status (monoparental vs biparental), mother and father age at birth of first child and the years of schooling of the mother and father. doi:10.1371/journal.pone.0071691.t003

t(18) = 20.81, P = 0.43

t(18) = 21.62, P = 0.87

3.262.22 Attention deficit (6 to 15 years)


Neuroendocrine-immunological abnormalities that are established during a stressful childhood are thought to mediate the development of the pro-inflammatory phenotype in adulthood [33,34] including reduced cortisol awakening response [86]. This could result in altered expression of certain cytokine levels in plasma, since cortisol levels are known to regulate immune and inflammatory responses. For example, stress has been shown to promote the expression of certain cytokines [87], correlations were found between reduced cortisol and increased aggression in adolescents and young men [88–90], and maltreatment in childhood was shown to lead to low basal cortisol in association with conduct and aggressive disorders [91]. However, the associations between stress, cortisol, cytokines and aggression are complex since opposite results have also been reported. Indeed, glucocorticoids produced by the adrenal gland are strong immunosupressors known to directly repress the expression of many cytokines in leucocytes including the cytokines examined here. For example, a study of the males from the present study during adolescence, reported that high cortisol levels were associated with high levels of aggression [92]. Therefore, both hyper- and hypo-active stress responses have been shown to associate with aggression [93]. Unfortunately, although our cohort was longitudinal and behavioral data was longitudinally collected from childhood, DNA from blood cells were collected only once the participants in the study reached adulthood. Future longitudinal studies should address these questions. Second, the results are based on a small sample of chronic aggressive subjects. The two longitudinal studies we used to recruit our subjects from had followed more than 1 000 males from childhood to adolescence. Young adult Caucasian males with a history of chronic physical aggression during childhood are relatively rare [11] and even more difficult to recruit for biological sampling. Thus, the limited numbers are an inherent property of the trait that we are examining. However replications of the present study with other longitudinal samples are obviously needed and should address this issue in part although we don’t expect a large number of CPA subjects in these future studies as well. In any case, the subgroup that was available for the DNA methylation analysis did not differ in its basic behavioral and socioeconomic characteristics from the original group (Table 3). Third, although we ruled out several of the most common confounders, there are some confounders that are tightly correlated with aggression. Cytokine levels as well as DNA methylation were previously associated with psychiatric diseases such as major depression [94], nevertheless the CPA and control groups were not significantly different in levels of anxiety and presence of psychiatric diagnoses as well as age. However, the CPA and control groups were significantly different with respect to variables that are also known to be strongly associated with chronic physical aggression trajectories from childhood to adolescence: childhood hyperactivity and opposition and adulthood criminal behavior [6,11]. Thus, it is impossible to determine whether the differential DNA methylated regions that we identified associate with aggressiveness or hyperactivity and criminal behavior. This however might be reflecting the simple fact that these behaviors are molecularly and functionally linked within the same biological pathways.

Conclusions This study has several implications. Taken together the data is consistent with the hypothesis that cytokines are involved in chronic physical aggression, hence that a peripheral immune component may play a key role in regulating these behavioral states. The results suggest that in addition to differential

CE Include

t(30) = 20.86, P = 0.40 (8) 4.0061.89 (7) 4.0062.00 (12) 3.2362.18

3.3661.93

(25)

t(18) = 22.93, P = 0.009 t(6.2) = 22.35, P = 0.06

t(30) = 20.09, P = 0.93 (8)

(8) 8.1466.71

3.4861.45 (7)

(7) 8.4367.00

3.4361.51 (12)

(12) 2.1961.92

3.3561.81

2.1661.86 Oppositional disorder (6 to 15 years)

Anxiety (6 to 15 years)

(25)

1.2060.96 Hyperactivity (6 to 15 years)

(25)

t(18) = 23.11, P = 0.006

F exact, 2 tailled: 0.10

t(30) = 22.79, P = 0.009

F exact, 2 tailled: 0.02 (4/7)

(8) 2.4761.20

57% (4/7)

(7) 2.4361.27

57% (1/10)

(12) 0.9161.03

10% 10% Self reported violence (21 years)

(25)

20% Criminal record (21 years)

(2/20)

F exact, 2 tailled: 0.64

F exact, 2 tailled: 0.019 F exact, 2 tailled: 0.02 (6/8) 75% (5/7) 71% (2/12) 17%

t(17) = 21.06, P = 0.31

Psychiatric record (21 years)

(5/25)

F exact, 2 tailled: 0.67 (3/7) 43% (3/7) 43% (6/10) 60% 33%

Familial adversityCE

(7/21)

t(18) = 20.26, P = 0.80 t(30) = 0.37, P = 0.71

t(29) = 20.73, P = 0.47 (7)

(8) 25.862.87

0.5160.41 (7)

(7) 25.2862.75

0.4460.42 (12)

(12) 25.462.71 25.6462.08

0.3460.30

Age at blood perfusion

(25)

Sample 2

0.3460.29

Sample 1

Sample 2 Sample 1

Mean ± SD or % (n)

(25)

Sample 2 Sample 1

Group effect CPA

Mean ± SD or % (n)

Control Variables

Table 3. Characteristics of the chronic physical aggression (CPA) group and control group in the original sample (1) and the selected sample (2) for the methylation analysis.


13



age at the birth of their first child and marital status when the subjects were age 6 [16,96]. Physical aggression and other behavior problems. In the course of the longitudinal studies, teachers annually rated the frequency of boys’ physical aggression, opposition, hyperactivity, inattention and anxiety from kindergarten to secondary school with the Social Behavior Questionnaire [96]. The physical aggression ratings were used to trace the developmental trajectories and create the CPA group and the control group (see [6] and [11] for details of the trajectory analyses). Self reported violence. During the data collection at 21 years, subjects were asked how often in the past year they had been implicated in physical fights and how often they had physically attacked someone. Criminal record. Canadian youth between 13 and 17 years who commit delinquent acts are referred to the juvenile courts. Subjects who were found guilty by a juvenile court were identified from official records. Using official records we also identified subjects in each group who had been convicted of a criminal offence between 18 and 24 years. Mental disorders. When the subjects were 15 years, structured psychiatric interviews using a French translation of the Diagnostic Interview Schedule for Children-2.25 (DISC-2.25) [97] were used with the mothers and the subjects to estimate the prevalence of DSM-III-R diagnoses such as: simple phobia, anxiety of separation, generalized anxiety, hyper anxiety, major depression, dysthymia, oppositional disorder and conduct disorder, over the previous 6 months [87]. Subjects were also asked whether they had a psychiatric record during the interview at 21 years.

methylation of critical regions within the cytokine gene loci, differential DNA methylation states of transcription factors that regulate these cytokine genes could be one of the mechanism that explain the decrease in plasma cytokines expression observed in the CPA group.

Materials and Methods Participants The subjects for the plasma cytokine level analysis (sample 1) [32] were recruited from male participants in two longitudinal studies of child development [15,95] which identified four different trajectories of physical aggression from age 6 to 15 years: Males in the first trajectory showed very low or no physical aggression at any time-point (low-never); males in the second trajectory started with low-to-moderate rates of physical aggression but declined over time (moderate level desister); those in the third trajectory began with high rates of physical aggression but subsequently declined between ages 12 and 15 (high level desister); males on the fourth trajectory showed consistently high physical aggression from age 6 to 15 years (chronic physical aggression, CPA). The two groups of Caucasian males we recruited from these longitudinal studies were all born in families with a low socioeconomic status and were living within 200 km from our laboratory at the time of recruitment. The first group was recruited from the chronic physical aggression trajectory (chronic physical aggression group, CPA, n = 8). The control group was recruited from the other three developmental trajectories (Control group, CG, n = 25). Quantification of 10 cytokine levels in plasma was done by ELISA for these subjects (sample 1) using the QuantibodyH Human Inflammation Array 1 from RayBiotech (Norcross, United States). The ELISA results showed that compared to the control group, the CPA group had significantly lower levels in plasma of five cytokines; lower pro-inflammatory interleukins IL-1a (T(28.7) = 3.48, P = 0.002) and IL-6 (T(26.9) = 3.76, P = 0.001), lower anti-inflammatory interleukin IL-4 (T(27.1) = 4.91, P = 0.00004) and IL-10 (T(29.8) = 2.84, P = 0.008) and lower chemokine IL-8 (T(26) = 3.69, P = 0.001) (see [32] for details). From these subjects (sample 1), the methylation analysis of these cytokine gene loci and their transcription factors was done on all the 8 subjects from the CPA group and a representative subgroup of 12 subjects was selected for the control group (sample 2). Note that one CPA subject was removed from the ELISA analysis due to cytokine concentrations outside of the expected range for the standard curve but was included in the methylation analysis. A schematic overview of the recruitment is illustrated in Figure S7. Characteristics of the original samples (n = 32 sample 1) and the selected subgroups (n = 20 sample 2) are presented in Table 3.

CD3+ T cells and Monocytes DNA Preparation For the study, 20 ml of blood were drawn in EDTA coatedtubes for each subjects. PBMC (whole mononuclear cells from peripheral blood) and T cell isolation procedures were adapted from Current Protocols in Immunology (1997, sections 7.1 and 7.5.1–7.5.11). Briefly, PBMC isolation was done by centrifugation with Ficoll-Paque (GE healthcare, Mississauga, Canada) and washed twice with HBSS (Hanks balanced salt solution, GIBCO, Burlington, Canada). T cells were isolated from the PBMCs by immunomagnetic isolation using CD3 dynabeads (Invitrogen (Dynal Biotech), Burlington, Canada). The beads were washed 3 times and incubated with the PBMCs for 45 min on a rotator at 4uC. Coated CD3+ cells with the dynabeads were isolated using a strong magnet (Steam Cell Technology) and washed 5 times with PBS/FBS. CD3+ cells coated with the dynabeads were then frozen at 280uC until DNA extraction. To isolate monocyte cells, B cells were extracted away from the remaining PBMC with CD9 dynabeads (Invitrogen (Dynal Biotech), Burlington, Canada) following the same procedure as for T cells isolation and then frozen at 280uC until DNA extraction. Monocytes and T cells DNA was extracted with WizardH Genomic DNA Purification kit (Promega, Madison, United States) following the manufacturer’s protocol.

Ethics Statement After complete description of the study to the subject, all participants provided written informed consent. The study was carried out in accordance with the Declaration of Helsinki, and was approved by the research ethics committee of the University of Montreal pediatric hospital (St-Justine Hospital).

Methylated DNA Immunoprecipitation (MeDIP), Amplification and Labeling The MeDIP analysis was adapted from [98]. Briefly, 2 mg of each of the T cells and monocytes DNA were sonicated and methylated DNA was immunoprecipitated with 10 mg of anti5methylCytosine (Calbiochem, Billerica, United States). Prior to sonication, two control plasmids were added to the DNA (6rg each), an unmethylated GFP plasmid and an in vitro methylated Luciferase plasmid. The DNA-antibody complex was immuno-

Assessment of Subjects’ Familial Adversity, Behavior Problems, Psychiatric Diagnoses and Criminal Records Familial adversity. The index of family adversity is a composite score of the degree of adversity in families ranging from 0 to 1, which has been used regularly with these cohorts. The index includes parent’s level of education, type of employment, PLOS ONE | www.plosone.org

14



if the p-value of the correlation was at most 0.05 (uncorrected for multiple testing), the absolute value of the coefficient derived from the linear model used to calculate the p-value was at least 0.5 and the false discovery rates (FDR) of the region-level statistic was at most 0.2. Cytokine levels in plasma were obtained from previous work [32] for the same set of individuals from which DNA methylation profiles were derived. The identification of regulatory elements contained in the cytokine and TF loci shown in the figures and tables were obtained using the Encyclopedia of DNA Elements (ENCODE) tracks from the UCSC genome browser (http://genome.ucsc.edu/ ENCODE/). The transcription factors binding sites and DNase hypersensitive sites (cluster) shown were obtained from ChIP-seq data done on 24 different cell lines. The histone marks shown were also obtained from ChIP-seq data done on 8 different cell lines but we selected only two cell lines relevant to the lymphocytes analyzed in our study, Gm12878 and K562. The microarray data are available at http://www.ncbi.nlm.nih. gov/geo under the accession number GSE47193.

precipitated with 5 mg protein A and the methylated DNA was eluted with 150 ml of TE at 1.5% SDS. The input and bound fractions were then purified and validated by PCR analysis for two control genes, H19 (methylated control) and b-actin (unmethylated control) and the two added plasmid, GFP (unmethylated control) and Luciferase (in vitro methylated control) using the following primers: H19: forward (59-TTGGTGGAACACGCTGTGATCA-39), reverse (59-GAGCCGCACCAGGTCTTCAG-39); b-actin: forward (59-AGCCATAAAAGGCAACTTTCG-39), reverse (59CCAACGCCAAAACTCTCCC-39); GFP: forward (CCAACGCCAAAACTCTCCC), reverse (AGCCATAAAAGGCAACTTTCG); Luciferase: forward (59-AGAGATACGCCCTGGTTCC-39), reverse (59CCAACACCGGCATAAAGAA-39). The input and bound fractions were then amplified in triplicate using the Whole Genome Amplification kit (Sigma, Oakville, Canada). The amplified input and bound fractions were labeled for microarray hybridization with either Cy3-dUTP or Cy5-dUTP (Perkin Elmer, Montreal, Canada) respectively using the CGH labeling kit (Invitrogen, Burlington, Canada) following the manufacturer’s instructions.

Microarray Validation Bisulfite treatment and pyrosequencing. 1 mg of EcoRI digested DNA was subjected to bisulfite treatment as previously described [105]. Loci were selected for further analysis from the locations of probes whose normalized intensities were significantly different between the groups (see Microarray Analysis). PCR amplifications were performed in two steps: 25 ng of bisulfite DNA were used for the outside PCR and 1 ml of this DNA for the nested PCR. The outside and nested PCR were performed using HotStarTaq DNA polymerase (Qiagen, Germantown, United States). For pyrosequencing analysis, 25 ml of the nested bisulfitePCR products were processed according to the manufacturer’s standard protocol (Biotage, Charlotte, United States). Sequencing reactions were performed with a PyroMark Gold Q24 Reagent Kit (Qiagen (Biotage), Germantown, United States) according to the manufacturer’s instructions. The percentage methylation at each CpG site was calculated from the raw data by use of PyroMark Q24-CpG Software (Biotage, Charlotte, United States). The mean methylation of all CG sites analyzed in each region was than tested for correlation with the previously obtained cytokine levels in plasma using Pearson correlation.

Microarray Design, Hybridization, Scanning and Analysis A detailed description of the methods and analyses concerning microarrays used in this study were previously described [60]. Briefly, custom 8644 K tilling arrays (Agilent technologies) containing probes selected to tile all the cytokines and TF loci (,100 Kb) were manufactured by Agilent (Mississauga, Canada). Three replicate microarrays were hybridized and analyzed for each sample. Extracted probe intensities were analyzed using the R software environment for statistical computing [99]. Log-ratios of the bound (Cy5) and input (Cy3) microarray channel intensities were computed for each microarray and then microarrays were normalized to one another using quantile-normalization [100] under the assumption that all samples have identical overall methylation levels. Differential methylation between groups of samples was determined at both the probe and region levels (1000 bp regions) to ensure both statistical significance and biological relevance as previously described [60]. At the probe level, a modified t-statistic was computed for each probe corresponding to probe log-ratio differences between CPA and control groups using the ‘limma’ package [101] of Bioconductor [102]. Then, region-level methylation differences were calculated as enrichment of large positive or negative t-statistic values among the probes in each 1000 bp region partition of the loci tiled with probes using the Wilcoxon rank-sum test. A probe and the containing region were called differentially methylated if the p-value of the probe t-statistic was at most 0.05 (uncorrected for multiple testing), log2-fold change between the groups was at least 0.25, and the false discovery rates (FDR) of the region-level statistic was at most 0.2. False positives due to multiple testing are than controlled using the FDR [103] using the method of Benjamini and Hochberg [104]. Figures 1, 2, 3, 4, S1, S2, S3, S4, S5 and S6 were obtained using the UCSC genome browser (http://genome.ucsc.edu/). They depict whole-chromosome views of tracks composed of average normalized probe intensity differences between groups and Pearson’s correlation coefficient (r) between cytokine levels in plasma and normalized MeDIP microarray probe intensities. Similarly to methylation differences, region-level statistical significance was calculated as enrichment probe correlations close to 1 or 21 among the probes in each 1000 bp region partition of the loci tiled with probes using the Wilcoxon rank-sum test. A probe and the containing region were called correlated with cytokine levels PLOS ONE | www.plosone.org

Supporting Information Figure S1 DNA methylation differences between CPA

(n = 8) and control (n = 12) groups in pro-inflammatory cytokines IL-6 and IL-1a loci in T cells and monocytes. Expanded views from the UCSC genome browser of IL-6 (A) and IL-1a (B) loci located on chromosomes 7 and 2 are depicted. The first two tracks shows the average MeDIP probe fold differences (Log2) between chronic physical aggressive (CPA) and controls groups and the average Pearson correlation coefficient values calculated between the methylation levels of each probe estimated from the microarray and the plasma IL-6 (A) and IL-1a (B) levels obtained from the same subject (n = 20) in monocytes. The following tracks show the same set of results but those obtained from T cells. In black are probes that are more methylated and in gray are those that are less methylated in the CPA group. In red are probes whose methylation level correlate positively with the cytokine level in plasma and in green are those that correlated negatively. Highlighted in blue are regions significantly differentially methylated between the groups in T cells, in red in monocytes and in purple in both cell type. The next track (CpG island) shows the location of the CpG islands (CG frequency .0.6) 15



found in the IL-6 (A) and IL-1a (B) loci. The regulatory element from ENCODE identified in these regions (see methods) are shown in the additional tracks. First, shown with black lines, is the location of DNase hypersensitive clusters where black indicate strong signal and grey a weaker signal from ChIP-seq data in 24 cell lines. Second, is the location of transcription factors (TF) identified from ChIP-seq data in 24 cell lines where black indicate a strong and grey weaker signal occupancy. The last tracks, identified the level of enrichment of three histone marks determined from ChIP-seq assay, histone 3 lysine 4 tri- and mono-methylation as well as histone 3 lysine 27 acetylation in two cell lines, GM12878 (pink) and K562 (blue). (TIFF)

cells, in red in monocytes and in purple in both cell type. The next track (CpG island) shows the location of the CpG islands (CG frequency .0.6) found in the IL-8 loci. The regulatory element from ENCODE identified in these regions (see methods) are shown in the additional tracks. First, shown with black lines, is the location of DNase hypersensitive clusters where black indicate strong signal and grey a weaker signal from ChIP-seq data in 24 cell lines. Second, is the location of transcription factors (TF) identified from ChIP-seq data in 24 cell lines where black indicate a strong and grey weaker signal occupancy. The last tracks, identified the level of enrichment of three histone marks determined from ChIP-seq assay, histone 3 lysine 4 tri- and mono-methylation as well as histone 3 lysine 27 acetylation in two cell lines, GM12878 (pink) and K562 (blue). (TIFF)

Figure S2 DNA methylation differences between CPA

(n = 8) and control (n = 12) groups in anti-inflammatory cytokines IL-4 and IL-10 loci in monocytes and T cells. Expanded views from the UCSC genome browser of IL-4 (A) and IL-10 (B) loci located on chromosomes 5 and 1 are depicted. The first two tracks shows the average MeDIP probe fold differences (Log2) between chronic physical aggressive (CPA) and controls groups and the average Pearson correlation coefficient values calculated between the methylation levels of each probe estimated from the microarray and the plasma IL-4 (A) and IL-10 (B) levels obtained from the same subject (n = 20) in monocytes. The following tracks show the same set of results but those obtained from T cells. In black are probes that are more methylated and in gray are those that are less methylated in the CPA group. In red are probes whose methylation level correlate positively with the cytokine level in plasma and in green are those that correlated negatively. Highlighted in blue are regions significantly differentially methylated between the groups in T cells, in red in monocytes and in purple in both cell type. The next track (CpG island) shows the location of the CpG islands (CG frequency .0.6) found in the IL-4 (A) and IL-10 (B) loci. The regulatory element from ENCODE identified in these regions (see methods) are shown in the additional tracks. First, shown with black lines, is the location of DNase hypersensitive clusters where black indicate strong signal and grey a weaker signal from ChIP-seq data in 24 cell lines. Second, is the location of transcription factors (TF) identified from ChIP-seq data in 24 cell lines where black indicate a strong and grey weaker signal occupancy. The last tracks, identified the level of enrichment of three histone marks determined from ChIP-seq assay, histone 3 lysine 4 tri- and mono-methylation as well as histone 3 lysine 27 acetylation in two cell lines, GM12878 (pink) and K562 (blue). (TIFF)


(n = 8) and control (n = 12) groups in cytokine’s transcription factor NFkB1 in T cells and monocytes. Expanded view from the UCSC genome browser of NFkB1 locus located on chromosomes 4 is depicted. The first two tracks shows the average MeDIP probe fold differences (Log2) between chronic physical aggressive (CPA) and controls groups and the average Pearson correlation coefficient values calculated between the methylation levels of each probe estimated from the microarray and the plasma levels of the cytokines it regulates (IL-1a, IL-6, IL8 and IL-10) obtained from the same subject (n = 20) in monocytes. The following tracks show the same set of results but those obtained from T cells. In black are probes that are more methylated and in gray are those that are less methylated in the CPA group. In red are probes whose methylation level correlate positively with the cytokine level in plasma and in green are those that correlated negatively. Highlighted in blue are regions significantly differentially methylated between the groups in T cells, in red in monocytes and in purple in both cell type. The next track (CpG island) shows the location of the CpG islands (CG frequency .0.6) found in the NFkB1 loci. The regulatory element from ENCODE identified in these regions (see methods) are shown in the additional tracks. First, shown with black lines, is the location of DNase hypersensitive clusters where black indicate strong signal and grey a weaker signal from ChIP-seq data in 24 cell lines. Second, is the location of transcription factors (TF) identified from ChIP-seq data in 24 cell lines where black indicate a strong and grey weaker signal occupancy. The last tracks, identified the level of enrichment of three histone marks determined from ChIP-seq assay, histone 3 lysine 4 tri- and mono-methylation as well as histone 3 lysine 27 acetylation in two cell lines, GM12878 (pink) and K562 (blue). (TIFF)


(n = 8) and control (n = 12) groups in pro-inflammatory chemokine IL-8 locus in T cells and monocytes. Expanded view from the UCSC genome browser of IL-8 locus located on chromosomes 4 is depicted. The first two tracks shows the average MeDIP probe fold differences (Log2) between chronic physical aggressive (CPA) and controls groups and the average Pearson correlation coefficient values calculated between the methylation levels of each probe estimated from the microarray and the plasma IL-8 levels obtained from the same subject (n = 20) in monocytes. The following tracks show the same set of results but those obtained from T cells. In black are probes that are more methylated and in gray are those that are less methylated in the CPA group. In red are probes whose methylation level correlate positively with the cytokine level in plasma and in green are those that correlated negatively. Highlighted in blue are regions significantly differentially methylated between the groups in T PLOS ONE | www.plosone.org


(n = 8) and control (n = 12) groups in cytokine’s transcription factor NFAT5 in T cells and monocytes. Expanded view from the UCSC genome browser of NFAT5 locus located on chromosomes 16 is depicted. The first two tracks shows the average MeDIP probe fold differences (Log2) between chronic physical aggressive (CPA) and controls groups and the average Pearson correlation coefficient values calculated between the methylation levels of each probe estimated from the microarray and the plasma levels of the cytokines it regulates (IL-8 and IL-4) obtained from the same subject (n = 20) in monocytes. The following tracks show the same set of results but those obtained from T cells. In black are probes that are more methylated and in gray are those that are less methylated in the CPA group. In red are probes whose methylation level correlate 16



positively with the cytokine level in plasma and in green are those that correlated negatively. Highlighted in blue are regions significantly differentially methylated between the groups in T cells, in red in monocytes and in purple in both cell type. The next track (CpG island) shows the location of the CpG islands (CG frequency .0.6) found in the NFAT5 loci. The regulatory element from ENCODE identified in these regions (see methods) are shown in the additional tracks. First, shown with black lines, is the location of DNase hypersensitive clusters where black indicate strong signal and grey a weaker signal from ChIP-seq data in 24 cell lines. Second, is the location of transcription factors (TF) identified from ChIP-seq data in 24 cell lines where black indicate a strong and grey weaker signal occupancy. The last tracks, identified the level of enrichment of three histone marks determined from ChIP-seq assay, histone 3 lysine 4 tri- and mono-methylation as well as histone 3 lysine 27 acetylation in two cell lines, GM12878 (pink) and K562 (blue). (TIFF)

from ENCODE identified in these regions (see methods) are shown in the additional tracks. First, shown with black lines, is the location of DNase hypersensitive clusters where black indicate strong signal and grey a weaker signal from ChIP-seq data in 24 cell lines. Second, is the location of transcription factors (TF) identified from ChIP-seq data in 24 cell lines where black indicate a strong and grey weaker signal occupancy. The last tracks, identified the level of enrichment of three histone marks determined from ChIP-seq assay, histone 3 lysine 4 tri- and mono-methylation as well as histone 3 lysine 27 acetylation in two cell lines, GM12878 (pink) and K562 (blue). (TIFF) Figure S7 Schematic overview of the recruitment pro-

cess. (TIFF) Table S1 List of cell lines and symbols included in the transcription factor binding site data from ENCODE used in the figures generated from UCSC genome browser. (DOCX)


(n = 8) and control (n = 12) groups in cytokine’s transcription factor STAT6 in T cells and monocytes. Expanded view from the UCSC genome browser of STAT6 locus located on chromosomes 12 is depicted. The first two tracks shows the average MeDIP probe fold differences (Log2) between chronic physical aggressive (CPA) and controls groups and the average Pearson correlation coefficient values calculated between the methylation levels of each probe estimated from the microarray and the plasma levels of the cytokines it regulates (IL-4 and IL-10) obtained from the same subject (n = 20) in monocytes. The following tracks show the same set of results but those obtained from T cells. In black are probes that are more methylated and in gray are those that are less methylated in the CPA group. In red are probes whose methylation level correlate positively with the cytokine level in plasma and in green are those that correlated negatively. Highlighted in blue are regions significantly differentially methylated between the groups in T cells, in red in monocytes and in purple in both cell type. The next track (CpG island) shows the location of the CpG islands (CG frequency .0.6) found in the STAT6 loci. The regulatory element

Spreadsheet S1 List of differentially methylated probes predicted from the cytokine and TF microarrays in T cells and monocytes. (XLSX)

Acknowledgments The authors wish to thank all the participants as well as the staff of the University of Montreal Research Unit on Children’s Psychosocial Maladjustment (GRIP) for their valuable assistance, Marilyne Blain for technical support, and Charles-E´douard Gigue`re for helping with the data banks.

Author Contributions Conceived and designed the experiments: NP MJS MS RET. Performed the experiments: NP DC DW. Analyzed the data: NP MJS. Contributed reagents/materials/analysis tools: MH FV. Wrote the paper: NP MJS MS RET.

References 10. Kokko K, Puilkkinen L (2006) Aggression in childhood and long-term unemployment in adulthood: a cycle of maladaptation and some protective factors. Dev Psychol 36: 463–472. 11. Nagin D, Tremblay RE (1999) Trajectories of boys’ physical aggression, opposition, and hyperactivity on the path to physically violent and nonviolent juvenile delinquency. Child Dev 70: 1181–1196. 12. Seguin JR, Pihl RO, Harden PW, Tremblay RE, Boulerice B (1995) Cognitive and neuropsychological characteristics of physically aggressive boys. J Abnorm Psychol 104: 614–624. 13. Campbell SB, Spieker S, Vandergrift N, Belsky J, Burchinal M (2010) Predictors and sequelae of trajectories of physical aggression in school-age boys and girls. Dev Psychopathol 22: 133–150. 14. Cote SM, Boivin M, Nagin DS, Japel C, Xu Q, et al. (2007) The role of maternal education and nonmaternal care services in the prevention of children’s physical aggression problems. Arch Gen Psychiatry 64: 1305–1312. 15. Nagin DS, Tremblay RE (2001) Parental and early childhood predictors of persistent physical aggression in boys from kindergarten to high school. Arch Gen Psychiatry 58: 389–394. 16. Tremblay RE, Nagin DS, Seguin JR, Zoccolillo M, Zelazo PD, et al. (2004) Physical aggression during early childhood: trajectories and predictors. Pediatrics 114: e43–50. 17. Bauer ME, Wieck A, Lopes RP, Teixeira AL, Grassi-Oliveira R (2010) Interplay between neuroimmunoendocrine systems during post-traumatic stress disorder: a minireview. Neuroimmunomodulation 17: 192–195. 18. Dowlati Y, Herrmann N, Swardfager W, Liu H, Sham L, et al. (2010) A metaanalysis of cytokines in major depression. Biol Psychiatry 67: 446–457. 19. Groer MW, Morgan K (2007) Immune, health and endocrine characteristics of depressed postpartum mothers. Psychoneuroendocrinology 32: 133–139.

1. Krug EG, Mercy JA, Dahlberg LL, Zwi AB (2002) The world report on violence and health. Lancet 360: 1083–1088. 2. NICHD Early Child Care Research Network. (2004) Trajectories of physical aggression from toddlerhood to middle childhood: predictors, correlates, and outcomes. Monogr Soc Res Child Dev 69: vii, 1–129. 3. Alink LR, Mesman J, van Zeijl J, Stolk MN, Juffer F, et al. (2006) The early childhood aggression curve: development of physical aggression in 10- to 50month-old children. Child Dev 77: 954–966. 4. Cote SM, Vaillancourt T, LeBlanc JC, Nagin DS, Tremblay RE (2006) The development of physical aggression from toddlerhood to pre-adolescence: a nation wide longitudinal study of Canadian children. J Abnorm Child Psychol 34: 71–85. 5. Tremblay RE (2010) Developmental origins of disruptive behaviour problems: the ‘original sin’ hypothesis, epigenetics and their consequences for prevention. J Child Psychol Psychiatry 51: 341–367. 6. Broidy LM, Nagin DS, Tremblay RE, Bates JE, Brame B, et al. (2003) Developmental trajectories of childhood disruptive behaviors and adolescent delinquency: a six-site, cross-national study. Dev Psychol 39: 222–245. 7. Campbell SB, Spieker S, Burchinal M, Poe MD (2006) Trajectories of aggression from toddlerhood to age 9 predict academic and social functioning through age 12. J Child Psychol Psychiatry 47: 791–800. 8. Barker ED, Boivin M, Brendgen M, Fontaine N, Arseneault L, et al. (2008) Predictive validity and early predictors of peer-victimization trajectories in preschool. Arch Gen Psychiatry 65: 1185–1192. 9. Barker ED, Seguin JR, White HR, Bates ME, Lacourse E, et al. (2007) Developmental trajectories of male physical violence and theft: relations to neurocognitive performance. Arch Gen Psychiatry 64: 592–599.


17



49. Jirtle RL, Skinner MK (2007) Environmental epigenomics and disease susceptibility. Nat Rev Genet 8: 253–262. 50. MacLennan NK, James SJ, Melnyk S, Piroozi A, Jernigan S, et al. (2004) Uteroplacental insufficiency alters DNA methylation, one-carbon metabolism, and histone acetylation in IUGR rats. Physiol Genomics 18: 43–50. 51. Ke X, Lei Q, James SJ, Kelleher SL, Melnyk S, et al. (2006) Uteroplacental insufficiency affects epigenetic determinants of chromatin structure in brains of neonatal and juvenile IUGR rats. Physiol Genomics 25: 16–28. 52. Weaver IC, Diorio J, Seckl JR, Szyf M, Meaney MJ (2004) Early environmental regulation of hippocampal glucocorticoid receptor gene expression: characterization of intracellular mediators and potential genomic target sites. Ann N Y Acad Sci 1024: 182–212. 53. Sinclair KD, Allegrucci C, Singh R, Gardner DS, Sebastian S, et al. (2007) DNA methylation, insulin resistance, and blood pressure in offspring determined by maternal periconceptional B vitamin and methionine status. Proc Natl Acad Sci U S A 104: 19351–19356. 54. McGowan PO, Suderman M, Sasaki A, Huang TC, Hallett M, et al. (2011) Broad epigenetic signature of maternal care in the brain of adult rats. PLoS One 6: e14739. 55. Terry MB, Ferris JS, Pilsner R, Flom JD, Tehranifar P, et al. (2008) Genomic DNA methylation among women in a multiethnic New York City birth cohort. Cancer Epidemiol Biomarkers Prev 17: 2306–2310. 56. Borghol N, Lornage J, Blachere T, Sophie Garret A, Lefevre A (2006) Epigenetic status of the H19 locus in human oocytes following in vitro maturation. Genomics 87: 417–426. 57. Heijmans BT, Tobi EW, Stein AD, Putter H, Blauw GJ, et al. (2008) Persistent epigenetic differences associated with prenatal exposure to famine in humans. Proc Natl Acad Sci U S A 105: 17046–17049. 58. Oberlander TF, Weinberg J, Papsdorf M, Grunau R, Misri S, et al. (2008) Prenatal exposure to maternal depression, neonatal methylation of human glucocorticoid receptor gene (NR3C1) and infant cortisol stress responses. Epigenetics 3: 97–106. 59. McGowan PO, Sasaki A, D’Alessio AC, Dymov S, Labonte B, et al. (2009) Epigenetic regulation of the glucocorticoid receptor in human brain associates with childhood abuse. Nat Neurosci 12: 342–348. 60. Borghol N, Suderman M, McArdle W, Racine A, Hallett M, et al. (2012) Associations with early-life socio-economic position in adult DNA methylation. Int J Epidemiol 41: 62–74. 61. Kinnally EL, Feinberg C, Kim D, Ferguson K, Leibel R, et al. (2011) DNA methylation as a risk factor in the effects of early life stress. Brain Behav Immun. 62. Uddin M, Galea S, Chang SC, Aiello AE, Wildman DE, et al. (2011) Gene expression and methylation signatures of MAN2C1 are associated with PTSD. Dis Markers 30: 111–121. 63. Uddin M, Koenen KC, Aiello AE, Wildman DE, de los Santos R, et al. (2011) Epigenetic and inflammatory marker profiles associated with depression in a community-based epidemiologic sample. Psychol Med 41: 997–1007. 64. Wang D, Szyf M, Benkelfat C, Provençal N, Turecki G, et al. (2012) Peripheral SLC6A4 DNA Methylation is Associated with In Vivo Measures of Human Brain Serotonin Synthesis and Childhood Physical Aggression. PLoS ONE In press. 65. Essex MJ, Thomas Boyce W, Hertzman C, Lam LL, Armstrong JM, et al. (2011) Epigenetic Vestiges of Early Developmental Adversity: Childhood Stress Exposure and DNA Methylation in Adolescence. Child Dev. 66. Naumova OY, Lee M, Koposov R, Szyf M, Dozier M, et al. (2012) Differential patterns of whole-genome DNA methylation in institutionalized children and children raised by their biological parents. Dev Psychopathol 24: 143–155. 67. Klengel T, Mehta D, Anacker C, Rex-Haffner M, Pruessner JC, et al. (2013) Allele-specific FKBP5 DNA demethylation mediates gene-childhood trauma interactions. Nat Neurosci 16: 33–41. 68. Provençal N, Suderman JM, Guillemin C, Massart R, Ruggiero A, et al. (2013) The signature of maternal rearing in the methylome in rhesus macaque prefontal cortex and T cells. J Neurosci. In press. 69. Uddin M, Aiello AE, Wildman DE, Koenen KC, Pawelec G, et al. (2010) Epigenetic and immune function profiles associated with posttraumatic stress disorder. Proc Natl Acad Sci U S A 107: 9470–9475. 70. Uddin M, Aiello AE, Wildman DE, Koenen KC, Pawelec G, et al. (2011) Epigenetic and immune function profiles associated with posttraumatic stress disorder. Proc Natl Acad Sci U S A 107: 9470–9475. 71. Pekowska A, Benoukraf T, Zacarias-Cabeza J, Belhocine M, Koch F, et al. (2011) H3K4 tri-methylation provides an epigenetic signature of active enhancers. EMBO J 30: 4198–4210. 72. Bogdanovic O, Fernandez-Minan A, Tena JJ, de Lacalle-Mustienes E, Hidalgo C, et al. (2012) Dynamics of enhancer chromatin signatures mark the transition from pluripotency to cell specification during embryogenesis. Genome Res. 73. Carotta S, Wu L, Nutt SL (2010) Surprising new roles for PU.1 in the adaptive immune response. Immunol Rev 238: 63–75. 74. Mostafa Anower AK, Shim JA, Choi B, Sohn S (2012) Pretreatment with interleukin-6 small interfering RNA can improve the survival rate of polymicrobial cecal ligation and puncture mice by down regulating interleukin-6 production. Eur J Pharmacol 688: 76–83. 75. Samuel JM, Kelberman D, Smith AJ, Humphries SE, Woo P (2008) Identification of a novel regulatory region in the interleukin-6 gene promoter. Cytokine 42: 256–264.

20. Hoge EA, Brandstetter K, Moshier S, Pollack MH, Wong KK, et al. (2009) Broad spectrum of cytokine abnormalities in panic disorder and posttraumatic stress disorder. Depress Anxiety 26: 447–455. 21. Janelidze S, Mattei D, Westrin A, Traskman-Bendz L, Brundin L (2010) Cytokine levels in the blood may distinguish suicide attempters from depressed patients. Brain Behav Immun 25: 335–339. 22. Kawamura A, Yoshikawa T, Takahashi T, Hayashi T, Takahashi E, et al. (2001) Randomized trial of phosphodiesterase inhibitors versus catecholamines in patients with acutely decompensated heart failure. Jpn Circ J 65: 858–862. 23. Koo JW, Duman RS (2008) IL-1beta is an essential mediator of the antineurogenic and anhedonic effects of stress. Proc Natl Acad Sci U S A 105: 751–756. 24. Misener VL, Gomez L, Wigg KG, Luca P, King N, et al. (2008) Cytokine Genes TNF, IL1A, IL1B, IL6, IL1RN and IL10, and childhood-onset mood disorders. Neuropsychobiology 58: 71–80. 25. O’Brien SM, Scott LV, Dinan TG (2004) Cytokines: abnormalities in major depression and implications for pharmacological treatment. Hum Psychopharmacol 19: 397–403. 26. Pace TW, Hu F, Miller AH (2007) Cytokine-effects on glucocorticoid receptor function: relevance to glucocorticoid resistance and the pathophysiology and treatment of major depression. Brain Behav Immun 21: 9–19. 27. Smith AK, Conneely KN, Kilaru V, Mercer KB, Weiss TE, et al. (2011) Differential immune system DNA methylation and cytokine regulation in posttraumatic stress disorder. Am J Med Genet B Neuropsychiatr Genet 156B: 700–708. 28. von Kanel R, Hepp U, Kraemer B, Traber R, Keel M, et al. (2007) Evidence for low-grade systemic proinflammatory activity in patients with posttraumatic stress disorder. J Psychiatr Res 41: 744–752. 29. Suarez EC, Lewis JG, Kuhn C (2002) The relation of aggression, hostility, and anger to lipopolysaccharide-stimulated tumor necrosis factor (TNF)-alpha by blood monocytes from normal men. Brain Behav Immun 16: 675–684. 30. Carpenter LL, Gawuga CE, Tyrka AR, Lee JK, Anderson GM, et al. (2011) Association between plasma IL-6 response to acute stress and early-life adversity in healthy adults. Neuropsychopharmacology 35: 2617–2623. 31. Alleva E, Cirulli F, Bianchi M, Bondiolotti GP, Chiarotti F, et al. (1998) Behavioural characterization of interleukin-6 overexpressing or deficient mice during agonistic encounters. Eur J Neurosci 10: 3664–3672. 32. Provençal N, Suderman MJ, Vitaro F, Szyf M, Tremblay RE (2013) Childhood chronic physical aggression associate with adult cytokine levels in plasma. PLoS One. In press. 33. Razin A, Szyf M (1984) DNA methylation patterns. Formation and function. Biochim Biophys Acta 782: 331–342. 34. Avni O, Rao A (2000) T cell differentiation: a mechanistic view. Curr Opin Immunol 12: 654–659. 35. Lee GR, Kim ST, Spilianakis CG, Fields PE, Flavell RA (2006) T helper cell differentiation: regulation by cis elements and epigenetics. Immunity 24: 369– 379. 36. Ansel KM, Djuretic I, Tanasa B, Rao A (2006) Regulation of Th2 differentiation and Il4 locus accessibility. Annu Rev Immunol 24: 607–656. 37. van Rietschoten JG, Gal-Yam EN, Jeong S, Cortez CC, Verweij CL, et al. (2008) Epigenetic regulation and nucleosome positioning in the human TATAless IL-1 alpha promoter. Genes Immun 9: 582–590. 38. Nile CJ, Read RC, Akil M, Duff GW, Wilson AG (2008) Methylation status of a single CpG site in the IL6 promoter is related to IL6 messenger RNA levels and rheumatoid arthritis. Arthritis Rheum 58: 2686–2693. 39. Andia DC, de Oliveira NF, Casarin RC, Casati MZ, Line SR, et al. (2010) DNA methylation status of the IL8 gene promoter in aggressive periodontitis. J Periodontol 81: 1336–1341. 40. Larsson L, Thorbert-Mros S, Rymo L, Berglundh T (2012) Influence of epigenetic modifications of the interleukin-10 promoter on IL10 gene expression. Eur J Oral Sci 120: 14–20. 41. Dalski A, Schwinger E, Zuhlke C (2001) Genomic organization of the human NFAT5 gene: exon-intron structure of the 14-kb transcript and CpG-island analysis of the promoter region. Cytogenet Cell Genet 93: 239–241. 42. Kim EG, Shin HJ, Lee CG, Park HY, Kim YK, et al. (2010) DNA methylation and not allelic variation regulates STAT6 expression in human T cells. Clin Exp Med 10: 143–152. 43. Li J, Chen F, Zheng M, Zhu H, Zhao D, et al. (2010) Inhibition of STAT1 methylation is involved in the resistance of hepatitis B virus to Interferon alpha. Antiviral Res 85: 463–469. 44. Fields PE, Kim ST, Flavell RA (2002) Cutting edge: changes in histone acetylation at the IL-4 and IFN-gamma loci accompany Th1/Th2 differentiation. J Immunol 169: 647–650. 45. Lee DU, Agarwal S, Rao A (2002) Th2 lineage commitment and efficient IL-4 production involves extended demethylation of the IL-4 gene. Immunity 16: 649–660. 46. Szyf M (2011) The early life social environment and DNA methylation: DNA methylation mediating the long-term impact of social environments early in life. Epigenetics 6: 971–978. 47. Waterland RA, Jirtle RL (2003) Transposable elements: targets for early nutritional effects on epigenetic gene regulation. Mol Cell Biol 23: 5293–5300. 48. Dolinoy DC, Weidman JR, Waterland RA, Jirtle RL (2006) Maternal genistein alters coat color and protects Avy mouse offspring from obesity by modifying the fetal epigenome. Environ Health Perspect 114: 567–572.


18



76. Jones EA, Flavell RA (2005) Distal enhancer elements transcribe intergenic RNA in the IL-10 family gene cluster. J Immunol 175: 7437–7446. 77. Hess J, Angel P, Schorpp-Kistner M (2004) AP-1 subunits: quarrel and harmony among siblings. J Cell Sci 117: 5965–5973. 78. Zalcman SS, Siegel A (2006) The neurobiology of aggression and rage: role of cytokines. Brain Behav Immun 20: 507–514. 79. Siegel A, Bhatt S, Bhatt R, Zalcman SS (2007) The neurobiological bases for development of pharmacological treatments of aggressive disorders. Curr Neuropharmacol 5: 135–147. 80. Nelson RJ, Chiavegatto S (2001) Molecular basis of aggression. Trends Neurosci 24: 713–719. 81. Lee DU, Rao A (2004) Molecular analysis of a locus control region in the T helper 2 cytokine gene cluster: a target for STAT6 but not GATA3. Proc Natl Acad Sci U S A 101: 16010–16015. 82. Chapoval S, Dasgupta P, Dorsey NJ, Keegan AD (2010) Regulation of the T helper cell type 2 (Th2)/T regulatory cell (Treg) balance by IL-4 and STAT6. J Leukoc Biol 87: 1011–1018. 83. Ando T, Nishimura M, Oka Y (2000) Decitabine (5-Aza-29-deoxycytidine) decreased DNA methylation and expression of MDR-1 gene in K562/ADM cells. Leukemia 14: 1915–1920. 84. Bartolomei MS (2009) Genomic imprinting: employing and avoiding epigenetic processes. Genes Dev 23: 2124–2133. 85. Widom CS (1989) Child abuse, neglect, and adult behavior: research design and findings on criminality, violence, and child abuse. Am J Orthopsychiatry 59: 355–367. 86. Platje E, Jansen LM, Raine A, Branje SJ, Doreleijers TA, et al. (2013) Longitudinal associations in adolescence between cortisol and persistent aggressive or rule-breaking behavior. Biol Psychol 93: 132–137. 87. Steptoe A, Hamer M, Chida Y (2007) The effects of acute psychological stress on circulating inflammatory factors in humans: a review and meta-analysis. Brain Behav Immun 21: 901–912. 88. Loney BR, Butler MA, Lima EN, Counts CA, Eckel LA (2006) The relation between salivary cortisol, callous-unemotional traits, and conduct problems in an adolescent non-referred sample. J Child Psychol Psychiatry 47: 30–36. 89. Popma A, Vermeiren R, Geluk CA, Rinne T, van den Brink W, et al. (2007) Cortisol moderates the relationship between testosterone and aggression in delinquent male adolescents. Biol Psychiatry 61: 405–411. 90. Shirtcliff EA, Granger DA, Booth A, Johnson D (2005) Low salivary cortisol levels and externalizing behavior problems in youth. Dev Psychopathol 17: 167–184. 91. Tarullo AR, Gunnar MR (2006) Child maltreatment and the developing HPA axis. Horm Behav 50: 632–639. 92. van Bokhoven I, Van Goozen SH, van Engeland H, Schaal B, Arseneault L, et al. (2005) Salivary cortisol and aggression in a population-based longitudinal study of adolescent males. J Neural Transm 112: 1083–1096.


93. Veenema AH (2009) Early life stress, the development of aggression and neuroendocrine and neurobiological correlates: what can we learn from animal models? Front Neuroendocrinol 30: 497–518. 94. Alesci S, Martinez PE, Kelkar S, Ilias I, Ronsaville DS, et al. (2005) Major depression is associated with significant diurnal elevations in plasma interleukin-6 levels, a shift of its circadian rhythm, and loss of physiological complexity in its secretion: clinical implications. J Clin Endocrinol Metab 90: 2522–2530. 95. Pingault JB, Tremblay RE, Vitaro F, Carbonneau R, Genolini C, et al. (2011) Childhood Trajectories of Inattention and Hyperactivity and Prediction of Educational Attainment in Early Adulthood: A 16-Year Longitudinal Population-Based Study. Am J Psychiatry. 96. Tremblay RE, Loeber R, Gagnon C, Charlebois P, Larivee S, et al. (1991) Disruptive boys with stable and unstable high fighting behavior patterns during junior elementary school. J Abnorm Child Psychol 19: 285–300. 97. Romano E, Tremblay RE, Vitaro F, Zoccolillo M, Pagani L (2001) Prevalence of psychiatric diagnoses and the role of perceived impairment: findings from an adolescent community sample. J Child Psychol Psychiatry 42: 451–461. 98. Keshet I, Schlesinger Y, Farkash S, Rand E, Hecht M, et al. (2006) Evidence for an instructive mechanism of de novo methylation in cancer cells. Nat Genet 38: 149–153. 99. R Development Core Team (2007) R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. 100. Bolstad BM, Irizarry RA, Astrand M, Speed TP (2003) A comparison of normalization methods for high density oligonucleotide array data based on variance and bias. Bioinformatics 19: 185–193. 101. Smyth GK (2005) Limma: linear models for microarray data. In: R. Gentleman VC, S Dudoit, R Irizarry, W Huber editor. Bioinformatics and Computational Biology Solutions using R and Bioconductor. New York: Springer,. 397–420. 102. Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, et al. (2004) Bioconductor: open software development for computational biology and bioinformatics. Genome Biol 5: R80. 103. Soric B (1989) Statistical ‘‘Discoveries’’ and Effect-Size Estimation. Journal of the American Statistical Association 84: 608–610. 104. Benjamini Y, Hochberg Y (1995) Controlling the False Discovery Rate - a Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society Series B-Methodological 57: 289–300. 105. Clark SJ, Harrison J, Paul CL, Frommer M (1994) High sensitivity mapping of methylated cytosines. Nucleic Acids Res 22: 2990–2997. 106. Armario A, Hernandez J, Bluethmann H, Hidalgo J (1998) IL-6 deficiency leads to increased emotionality in mice: evidence in transgenic mice carrying a null mutation for IL-6. J Neuroimmunol 92: 160–169.

19
